Add SATA SSD to NVMe Volume For Mirroring?

panzerscope

Contributor
Joined
May 30, 2022
Messages
146
Hello all,

Quick question. I currently have a pool created on my NVMe drive which hosts a variety of Windows VM's. At the time of creating the volume, this was setup as a single drive only with no mirroring. For the sake of safety I would like to add drive to enable mirroring, just in case the NVMe ever goes down. However due to platform restriction with respect to PCIe lanes, I am unable to add in another NVMe.

My question is, can I instead a SATA 3 SSD of the same capacity to the volume to enable mirroring ?

Many thanks,
P
 

HoneyBadger

actually does care
Administrator
Moderator
iXsystems
Joined
Feb 6, 2014
Messages
5,112
Hello @panzerscope - you certainly can add a SATA SSD as a mirrored volume using the Storage -> Manage Devices -> Extend workflow in the SCALE webUI. Your overall pool speed may "slow down" if the SATA SSD isn't able to deliver the same IOPS/bandwidth, which you might feel at the VM level if your network is faster than the SATA drive - but there's no technological limitation here.
 

Arwen

MVP
Joined
May 17, 2014
Messages
3,611
Their are inexpensive PCIe adapter cards to M.2 NVMe card slots. If you have a free x4 lane, (or wider), PCIe slot you can add such a card.
 

drinking12many

Contributor
Joined
Apr 8, 2012
Messages
148
Yeah Id add a card before I did that you will likely take a hit on performance. There are quite a few that can support 2 nvme drives provided your motherboard can do bifurcation (Note: some say they can but cant really) So I have a Supermicro X9 series motherboard and no matter how I set up the bifurcation in the bios it would only see one nvme, I had enough slots I just bought another like 30 dollar card on amazon and just put one in each card. https://www.amazon.com/gp/product/B09PGDMWKH/ref=ppx_yo_dt_b_asin_title_o08_s00?ie=UTF8&psc=1
 

panzerscope

Contributor
Joined
May 30, 2022
Messages
146
Thanks guys. I will take a look at the current configuration I have in my system to see if I can add an additional NVMe for the mirroring as I would prefer to not take a hit on the performance that would come with a SATA SSD. Looks like the wonder of ZFS is that it is super flexible with adding to your arrays.
 

HoneyBadger

actually does care
Administrator
Moderator
iXsystems
Joined
Feb 6, 2014
Messages
5,112
Looks like the wonder of ZFS is that it is super flexible with adding to your arrays.

Where mirrors are concerned, yes - you can quite easily change the width of a stripe/mirror vdev both up and down, or even remove individual vdevs from an all-mirror pool.

The counterpoint to that is that there's presently no way to increase the width or redundancy of a RAIDZ vdev after creation; for example, if you decide that a 6-wide RAIDZ2 is where you're at with a 12-bay chassis, but later wish to change to a 16-bay and 8-wide Z2, you can't do that in-place. Some testing is underway for this at the OpenZFS level, but it's a complicated bit of code, and no one wants to lose their data.

That's one of the many reasons why mirror vdevs are recommended so frequently here.
 

panzerscope

Contributor
Joined
May 30, 2022
Messages
146
Where mirrors are concerned, yes - you can quite easily change the width of a stripe/mirror vdev both up and down, or even remove individual vdevs from an all-mirror pool.

The counterpoint to that is that there's presently no way to increase the width or redundancy of a RAIDZ vdev after creation; for example, if you decide that a 6-wide RAIDZ2 is where you're at with a 12-bay chassis, but later wish to change to a 16-bay and 8-wide Z2, you can't do that in-place. Some testing is underway for this at the OpenZFS level, but it's a complicated bit of code, and no one wants to lose their data.

That's one of the many reasons why mirror vdevs are recommended so frequently here.
Thanks for the insight, fingers crossed that eventually increasing the RAIDZ Vdev in the future will be possible. So far ZFS is really flexible, so I am sure that after some time they will be able to overcome this.
 
Top